Published in December 2003

PRIVACY IN THE OFFICE ENVIRONMENT
By Joel Lewitz, PE

Understanding the sound and the silence: applications of sound masking in open- and closed-plan environments, with possible HIPAA requirements.

Applications of sound masking to open-plan offices are well documented. Recent projects emphasize the application of sound masking to closed-plan environments. There are similarities and differences, which we’ll cover here.
     Of the many definitions of noise, the one that best applies to the office environment is, “unwanted sound.” A good example is the conversation I had with a facility manager who had just moved into an open-plan office designed with a sound-masking system. There had been a power failure and the sound-masking system went off. She said, “Suddenly it got real noisy.” I thought for a moment, because I knew that the background noise level had just dropped from about 47dBA to 37dBA. This is a subjective reduction of 10dB or about half as loud as when the sound masking was operational. Of course, she meant that she was suddenly aware of all the conversations from distant workstations. The intelligibility and distraction were the noise to her. The level of unwanted sound had gone way up.

How to Reduce Noise?
     How do we reduce noise and achieve privacy and freedom from distraction? Let’s review the basics of noise control. Applying one or more of these six techniques can alleviate most noise problems.
Elements of Noise Control:
1. Control the noise at the source.
2. Move the source and receiver farther apart.
3. Enclose the source.
4. Enclose the receiver.
5. Mask the signal at the receiver.
6. Active noise control (noise cancellation).
     For any noise problem, increasing the number of elements that are brought to bear will increase the success of the result. Trying to do the whole job with just one element will prove costly, ineffective, unacceptable or simply not feasible.
     Fortunately, in the office environment, typically we can apply the first five elements. If this happens and the work is done properly, we will be successful. The Talker-Path-Listener concept takes into account the major variables associated with speech privacy. In terms of our balanced approach to noise control and speech privacy, we can:
• Control the noise at the source by controlling the talker voice level (Element 1): There are limits to this element. Many consultants are using raised voice, rather than normal voice levels in their speech-privacy analysis. We also have to consider speakerphones or amplified sound in videoconference rooms as a source. If the person is a “shouter,” he will never have privacy unless extreme measures are taken, such as moving him to a soundproof office. This partially incorporates Elements 2 and 3, but is still rather impractical.
• Move the source and receiver farther apart, enclose the source, enclose the receiver. All apply to some form of space planning, architectural design, partition type and dimensions and materials selection in the source-to-receiver path (Elements 2, 3 and 4). There are limits to these elements, which often relate to cost, but because we are applying three of the six noise-control techniques, this is a very important part of the successful solution.
• Mask the signal at the receiver with masking sound: Privacy modeling software has become more elegant and powerful recently. What is evident from the modeling tools is how powerful an element sound masking really is. Furthermore, when the project is finished, and increased privacy is desired, it is far easier to increase the masking a few decibels than it is to increase the partition height a few feet.
     There are also limitations. We could make it sound like Niagara Falls in the office. The privacy would be perfect, but no one would want to be there, let alone try to work.
     Regarding Element 6, active noise cancellation cannot be applied to office privacy yet. We do not have the necessary computer-processing power at this time.
     Path considerations differ between the open- and closed-plan environments. These physical variables have costs associated with them that often limit their effectiveness. We worked on a project where the company president dictated that all offices would be open plan, including his own. A glance at the space plan revealed that the closest workstation to the president’s was 50 feet away in any direction. There was no question that he had confidential speech privacy. In this project, distance attenuation was used to achieve confidential privacy for one receiver. The cost of office space prohibits allocating 2500 square feet to each worker’s cubicle.

Open Plan Offices
     Variables associated with speech privacy in open-plan offices are well documented. Figure 1(pdf) shows the typical open plan and the schematic sound transmission path between talker and listener. Architect, space planner and facility managers have control over the variables that will determine the interzone attenuation of unwanted sound between talker and listener. These include:
• Distance between work stations
• Workstation partition height and width
• Workstation partition sound transmission class (STC)
• Ceiling reflection attenuation
• Vertical surface path attenuation.

Closed Plan Offices
     Concerns for speech privacy are not limited to the open-plan office environment. Congress, as part of a broad healthcare reform, enacted the Healthcare Insurance Portability and Accountability Act of 1996 (HIPAA), which addresses accountability for privacy in terms of protection of written or electronic healthcare information. Under the oral communications section, healthcare providers must also provide reasonable safeguards to ensure that conversations with or about a patient are private. The awareness of the need for speech privacy in healthcare facilities with enclosed offices such as exam rooms and doctor’s offices has increased in recent years.
     In the corporate environment, companies are moving into “spec” buildings, which typically have ceiling height partitions between offices. The “standard partition” usually cannot provide confidential privacy such as might be required for attorneys or HR-related activities.
     Figure 2 (pdf) shows the typical closed-plan sound transmission paths between talker and listener. These include:
• Ducted connections to supply air grilles
• Lightweight ceiling with open joints
• Recessed light fixtures
• Open plenum return air grilles
• Partition-to-ceiling joints
• Glazing and edge leaks at window mullions
• Recessed receptacles and light switches in the same stud cavity
• Gaps under base plate.
     As with the open-plan path variables, the closed-plan variables are numerous. Adding to the complexity is the importance of detailing the sound-rated construction. Penetrations in sound-rated construction, such as pipes, ducts, conduit and boxes, must be caulked and sealed airtight. Small cracks and gaps can allow a lot of sound to penetrate a perfectly good sound-rated partition, seriously compromising privacy.
     Figure 3 (pdf) illustrates four common partition conditions for enclosed offices and the approximate STC rating for each. [See also Table A.]
     Compare Figure 3 with the subjective privacy criteria in Figure 4 (pdf). The ceiling height partition with mineral fiber ceiling tile, at STC 36, is just barely into the low end of the moderate privacy range. This condition is commonly found in office buildings and healthcare facilities. If the quality of the construction is not perfect, the STC rating will deteriorate to below STC 35. Without sound masking, there will be no speech privacy. This condition will not be compliant with HIPAA guidelines unless sound masking is introduced into the design.

Privacy Criteria
     Once the source noise level has been attenuated over the path between talker and listener, it is the ratio of this signal to the background noise level that determines the degree of speech privacy. This signal-to-noise ratio is at the heart of both the subjective perception and objective measurement of speech privacy.
     The Articulation Index (AI) was developed in response to the need to quantify how much intelligibility is in a noisy environment. For example, this would be important in a noisy airplane cockpit, particularly if the pilot were trying to communicate with the control tower or crew. AI is computed by summing the contributions of the signal-to-noise ratio in each of 15 1/3 octave frequency bands from 200Hz to 5000Hz. Before the arithmetic sum is taken, the 15 signal-to-noise ratios are weighted against how much speech intelligibility is carried in that particular frequency band. The 1/3 octave band that is assigned the highest weighting is centered at 2000Hz. According to the standard, it carries the most intelligibility. Close behind is the 1600Hz 1/3 octave band.
     Real-world applicability of the standard can be demonstrated easily if you have a band-limiting filter through which you can play a voice recording. If you listen only to the frequencies between 1782Hz and 2245Hz (the 1/3 octave band centered at 2000Hz), it will sound like an old transistor radio from the 50s—but intelligibility will be fairly good. In the 1/3 octave band centered at 5000Hz, you will hear a few “s”s and “t”s, but you won’t understand a single word. Likewise, down at 200Hz, you will hear a few low booms, but there will be nothing that contributes to intelligibility of the speech recording.
     AI goes from 0 to 1, where zero is no articulation (full privacy) and one is full articulation (no privacy). In the early 60s, efforts were made to correlate AI’s objective measurement to subjective privacy criteria, as shown in Table B.
     Because the highest AI results in the least privacy, another related scale was required to give the higher privacy the higher value. This is the Privacy Index, or PI. PI is 100X(1-AI). Furthermore, the categories were reduced for simplicity. Today, the generally accepted speech privacy definitions are listed in Table C.

Masking System
     Figure 5 (pdf) illustrates a typical range for masking noise spectra. Most consultants use some form of this curve. A discussion of the characteristics of this curve will greatly enhance its application in the field.
     Because we are trying to mask speech, we need masking in the frequencies that carry most of the intelligibility. The AI weighting and other studies tell us that this is in the middle frequencies, from about 250Hz to 4000Hz. We don’t need energy below 250 and above 4000 to mask intelligibility because there is no intelligibility to mask. Energy above 4000Hz will only make the masking system sound hissy. (Figure 5 doesn’t have a lower limit above 4000Hz.)
     Below 250Hz and above 4000Hz, we want sufficient energy to create a smooth curve and make the sound more natural and acceptable. Figure 5 shows a masking spectrum down to 125Hz. This is the lowest practical capability of a cost-effective masking speaker. Some consultants’ masking spectrum goes down to the 63Hz octave band, but at that frequency there has to be some contribution from the mechanical system for the energy in that band.
     Another aspect of Figure 5 is the upper and lower limit. First, the requirement for sound masking will vary depending on the contribution to signal-to-noise ratio from other variables such as voice level and path attenuation. Also, the privacy criteria may vary. Nevertheless, practically speaking, masking systems typically are not set higher than about 51dBA. For enclosed offices, the upper limit would fall about 5dB to 46dBA because the office is a more confined space and the other elements, such as a ceiling height partition between talker and listener contribute more to the overall equation than a partial height screen such as found in cubicles.
     The dBA level does not tell the whole story. The spectrum must also be correct. We were called into a facility with a masking system that still had privacy problems. The masking system was operating at 46dBA, which should have been sufficient for this office. However, most of the 46dBA was above 2kHz. The rest of the spectrum was below the preferred masking curve. The speech frequencies lacked sufficient masking sound to achieve adequate privacy.

It’s Not White, Not Pink
     White noise is a random noise that contains an equal amount of energy per frequency band: that is, 100-200, 800-900, 3000-3100, etc.
     Pink noise, by definition, has an equal amount of energy per octave. The bands 0-200, 800-1600 and 3000-6000 all contain the same amount of energy. An octave band is a band of frequencies in which the upper band limit is nearly two times the lower band limit (within 2%). For example, the octave band centered at 1000Hz is roughly from 707Hz to 1414Hz. The octave band center frequencies we are familiar with are set by standard and are shown on the abscissa of Figure 6 (pdf).
     Pink Noise would be a straight line on the octave band graph paper in
Figure 6. White noise, when plotted on octave band paper, goes up 3dB per octave.
It’s not even an NC curve. Beranek’s NC curves and Blazier’s RC curves were intended to establish HVAC system design goals. We are attempting to follow mechanical system noise spectra for reasons of subjective acceptability, but we neither need nor want the higher frequency component of the NC curve that typically results from air flow through supply diffusers.
     Beware: Mimicking of the mechanical system can backfire. In one project, an open-plan office lacked privacy simply because it was too quiet. After a masking system was installed, we asked the facility manager if the occupants felt the privacy had improved. He said the privacy was excellent and there were no complaints about the masking system. However, one person complained that the office was too cold.

Conclusion
     Confidential privacy criteria for many sectors of business, commerce and healthcare organizations (see HIPAA at www.hipaa.org) are bringing sound masking requirements into more closed-plan office projects. Awareness and attention to the differences between the physical characteristics of the work environment and correct sound masking design, tuning, balance and adjustment for both open and closed plan will turn the sound into the silence successfully for privacy-sensitive environments.

Founding principal of Lewitz and Associates, Inc., Joel Lewitz has more than 30 years of experience in acoustics and sound-system design. A member of Sound & Communications’ Technical Council, he holds M.S.E. (Electrical Engineering) and B.S.E. (Electrical Engineering) degrees from the University of Michigan.